Home / The Selection Axis

The Selection Axis

The Selection axis is the framework's name for the domain side of any transformation step — the choice of what a step admits and what it excludes. It carries two of the four directional primitives: Toward (admit or increase coupling) and Away (exclude or decrease coupling). Where the companion Routing axis governs where the result goes, Selection governs what gets let in to begin with. Together the two axes are how the framework explains why there are exactly four primitives rather than some other number.

The carving idea

A central move of the framework is that the primitive set is not merely listed at four; it is carved at four. Listing four things is arbitrary; carving derives the count from structure. The carving claim is that a recursive transformation step has two formal sides — a domain side (Selection) and a codomain side (Routing) — each a binary choice. The product of the two binaries, {toward, away} × {loop-back, propagation}, yields the four primitives. Selection supplies the first factor.

This is where the contestability gradient bites: different parts of the claim are held with different firmness. The framework separates a no-fewer claim from a no-more claim, and treats them very differently.

Why no fewer than four

The two axes are independent: Selection cannot collapse into Routing, and Routing cannot collapse into Selection. The proof of independence is behavioral — the four combinations admit-and-retain, admit-and-emit, exclude-and-retain, exclude-and-emit are genuinely distinct in what they do. Because neither axis can be dropped without losing distinctions, the primitive set cannot shrink below four. This no-fewer claim is held as frame-internal (tautological): the distinctness follows by construction, so the only way to refuse it is to decline the carving itself.

The Selection/Routing decomposition is retained throughout the framework because it is load-bearing twice over: it generates both the emotion set (through predicate binding) and the Force set. Its usefulness is not in dispute.

Why "no more" is only contested

The harder claim is no more than four — that the two axes exhaust the elementary sides of a step. The framework's argument is genuinely strong as a mapping: magnitude, timing, medium, gain, threshold, and channel all describe how a step occurs and add no third directional side, and the hold/maintain stress test resolves cleanly. But the 2026 adversarial run surfaced candidate motions — for example a memory or temporal operation — that the two-axis carving does not obviously contain. So the exactly-four / no-more claim is held as a carving / count claim: apt and defensible, but contestable by counter-instance. The framework's own benchmark independently treats four-primitive sufficiency as an open proof-burden.

The honest position, then, is that the four-primitive set is a contestable carving — the most accurate carving available, not a closure exempt from challenge. As the framework's no-possession stance predicts, a carving can be the best available without being final.

Common misreadings

Selection is not a ranking or hierarchy of primitives, and it is not a sealed, challenge-proof closure. A specific historical error must be flagged: prior editions invented "emotion-generator" axes — {toward, away} × {relation, integrity} — and even a later {toward, away} × {relation, value-ledger} pair. There are no such axes in the corpus; they are withdrawn. The genuine domain-side axis is Selection, full stop.

Formal status

Formal status. Epistemic: Derived. The no-fewer claim is frame-internal — the four combinations are behaviorally distinct by construction, contestable only by declining the carving. The exactly-four / no-more claim is carving/count, contestable by counter-instance, and the benchmark treats four-primitive sufficiency as an open proof-burden. Alethic: the decomposition aspires to map the real structure of a transformation step and maps it well — the surrender is of closure, not of accuracy. Provenance: the carving is canonical (the four semantics are authority-fixed), but its exhaustiveness is treatise-side demoted to a contestable carving on a benchmark-flagged open burden.